The Lab Notes

Google

Gemini 4 Hydrogen costs twice Gemini 3.8 Flash until Dec. 31

Google made Gemini 4 Hydrogen generally available at $1.50 per 1M input tokens and $7.50 per 1M output, double the current Gemini 3.8 Flash rate.

A tilted clear glass gas cylinder with a brass valve and orange and yellow shapes inside, on a blue background patterned with rows of yellow half circles.

Google has made Gemini 4 Hydrogen, its model for fast everyday coding, generally available in the Gemini API at twice the current price of Gemini 3.8 Flash, the model Google tells developers to upgrade from. The two list at the same price once the Gemini 3.8 Flash rate rises.

Google released Hydrogen, model ID gemini-4-hydrogen, as generally available in the Gemini API on Nov. 12, according to the Gemini API release notes.

Hydrogen Costs $1.50 Per 1M Input Tokens

Google’s pricing page, read Nov. 15, lists Hydrogen at $1.50 per 1M input tokens and $7.50 per 1M output tokens on the paid tier. Output pricing includes thinking tokens, and the rate stays flat across the full 1M-token context window.

Gemini 3.8 Flash Reaches the Same Price on Jan. 1

The same pricing page lists Gemini 3.8 Flash at $0.75 input and $3.75 output through Dec. 31. From Jan. 1, 2027, it lists at $1.50 and $7.50, the same as Hydrogen. Google’s migration guide recommends Hydrogen as the upgrade from Gemini 3.8 Flash.

Hydrogen Is the Fast Model Beside the Frontier Pair

Google describes Hydrogen as built for speed and efficiency in everyday coding and agentic workflows. Its frontier models, Gemini 4 Carbon and Gemini 4 Argon, handle deep, long-horizon reasoning.

Google’s own figures, from its launch post on Nov. 12, compare Hydrogen with Argon and Gemini 3.8 Flash. On the Artificial Analysis Intelligence Index, Google said Hydrogen scores 51 at the default level, against 50 for Argon, and 54 at xhigh, its highest thinking setting. On Terminal-Bench 4.1, Google said Hydrogen scores 57%, against 18.2% for Gemini 3.8 Flash, and finishes tasks 2.5 times faster at its default setting.

Moving From Gemini 3.8 Flash Takes Code Changes

Hydrogen does not support explicit context caching, so developers must remove cached_content fields and stop calling cachedContents and caches.create. Implicit caching runs by default on prompts of at least 4,096 tokens and discounts cached input by 95%, to $0.075 per 1M tokens.

Thinking now starts at high rather than medium, and the minimal level returns an error. Developers should remove temperature, top_p and top_k, since in-range values are ignored and out-of-range values return a 400 error.

Hydrogen Rolls Out Across Google Products

Google began a global rollout on Nov. 12 in the Gemini API, Google AI Studio, Android Studio, Gemini for business and Google Search for all users. Gemini in Google Sheets starts with Google AI Pro and Ultra subscribers, and Antigravity has it for paid Pro and Ultra subscribers.

Our View

We think the date on the Gemini 3.8 Flash price matters more than Hydrogen’s own price. Until Dec. 31, switching costs twice the rate for the same number of tokens. From Jan. 1 the two models list at one rate, so staying on Gemini 3.8 Flash saves nothing and the newer model costs no more. Developers on Gemini 3.8 Flash have a short stretch of cheaper bills to weigh against the work of moving.

That work is mostly caching. Explicit caching means asking the API to store a long prompt and then referring to it on later requests. Hydrogen does not accept those requests, so every call that uses them has to be rewritten. Implicit caching, where the API recognizes a repeated prompt on its own, takes over, but only for prompts of 4,096 tokens or more. Shorter prompts fall outside that discount.

The switch pays off only if Hydrogen’s speed on a team’s own tasks comes close to the figures Google published, and only a test on that traffic can show it.